Camera-Based Localization in GNSS-Denied Mining Environments via Nonlinear Mapping
Year: 2026
Type: conference-paper
Abstract: Camera-based infrastructure localization is a promising solution for positioning in GNSS-denied mining environments. However, existing approaches commonly rely on homography-based pixel-to-world mapping under a single-plane assumption, which can lead to systematic localization errors on uneven terrain with slope changes and local elevation differences. To address this limitation, this paper proposes applying a nonlinear mapping network to the pixel-to-world mapping stage, allowing the relationship between image observations and two-dimensional ground positions in the world coordinate system to be learned directly under uneven mining-terrain conditions. The current field evaluation focuses on personnel localization in a real mining environment using synchronized fixed-camera observations and total station ground truth, and compares the proposed method directly with a homography-based baseline under identical detection inputs and a unified evaluation protocol. The results show that the two methods perform comparably on approximately planar ground, whereas the proposed nonlinear mapping achieves lower localization errors and better robustness on uneven terrain.
Institution: RWTH Aachen University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7225861701
FWCI:
Cites:
Cited by:
Related to: